NASA Technical Reports Server (NTRS) 20020039159: Multiagent Modeling and Simulation in Human-Robot Mission Operations Work System Design

10 Pages·0.61 MB·English

by  NASA Technical Reports Server (NTRS)

Multiagent Modeling and Simulation in Human-Robot
Mission Operations Work System Design
Maarten Sierhuis _ William J. Clancev 2,Michael H. Sims
Computational Sciences Division
,_]qSA/Arnes Research Center
Moffett Field. CA 94035
[msterhuis _ msims {_mail.arc.nasa. gov
Abstract proposed discovery mission to the Moon--the Victoria
This paper describes a collaborative multiagent Mission. We begin by describing our view of work
modeling and simulation approach for designing work practice and how it is modeled in Brahms. We then
systems. The Brahms environment is used to model describe Victoria, the model, and simulation results.
mission operations for a semi-autonomous robot
Z. Work Practice
mission to the _Woonat the work practice level. It shows
the impact of human-decision making on the activities The concept of work practice originates in the
and energy consumption ofa robot. research disciplines of socio-technical systems, business
A collaborative work systems design methodology is anthropology, and management science. Work systems
described that allows informal models, created with design, as presented here. has its roots in the design of
users and stakeholders, to be used as input to the socio-technical systems. This method was developed in
development offormal computational models. the 1950s by Eric Trist and Fred Emery [2], to
understand and leverage the advantages of the social and
1. Introduction
technical aspects of work. Work systems design extends
Work systems involve people, machines, tools, this tradition by focusing on both the informal and
documents, and facilities interacting in activities over formal features of work by applying ethnographic
time. These activities produce goods, services or--as is participant observation [3].
the case in the work system described in this paper-- A work practice is defined as the collective activities
scientific data. Many work systems we encounter of a group of people who collaborate and communicate,
everyday have existed over a long period of time. while performing these activities synchronously or
Improvement of such work systems is olden done asynchronously. In contrast, aTayioristic view of work
through business process analysis and reengineering [1]. abstracts circumstances, tools, and behaviors into an
But managers must also design work systems de novo. idealized transformation of work products by job
One of the challenges of work system design is that functions.
work systems are often large and complex and persist To describe how work actually gets done, we need to
over a tong period of time. This makes the design include those aspects of the situation that influence
process complex and non-deterministic. In this paper we individual acnviry (not only the procedures and
describe Brahms, a multiagent modeling and simulation problem-solving behavior), such as collaboration,
(M&S) environment for designing complex interactions conversations, "off-task" activities, muln-tasking,
in human-machine systems. interruption and resumption of work _and how this _s
In Brahms. we model work processes at the work managed), intbrmal interaction such as assistance
practice level. We describe a case study in which we outside ofjob function, use of tools, and t'actlittes [41
used Brahms to design mission operations for a
'Rc.w,zarch [nshtutc of_Mlvanccd Comput_ Science (RIACS_USRA}.
:Institute tbr Human and Machine Cognition I.Univctslty ofWcs¢ Florida, Pensacola_
l]r:thtn_, _,,.I 'vl&',; environment fur rcprescntmg work workti'ames (called ,h,tcutahh,_), thus ,Jb,,cr,,,ttton is
pr.lchcc ar,u_g 4 rnult_agent rule-based actlv_ty language, dependent oll what the agent Lsdoing
.'hat can hc swnulatcd using the Brahms ,;imulation Communication Modeh 4croons by _hich agents
cngmc [51. and objects exchange helicl_, inctudmg telling someone
This paper discusses how we have used Brahms to something or asking a questiun. A conversatton ts
design the work system tbr the proposed Victoria modeled as an activity with commumcation actsuns,
mtss,on. The attentive reader might question how we can etther thee-to-face or through some de',_ce. ,uch as a
desrgn a work practice'? Indeed. a work practice is not telephone or email. The choice of dex ice and how it is
designed, but emerges over time. However, our interest used are part of the work practice
t.,, to determine whether a model at the work practice Typically a Brahms model is _kctchcd b? specifying
le,,el can be used m the destgn of a mission. We believe the geography and groups first. The gram size of the
that a work practice model is useful, because it provides simulatton clock (time per tickj may vary from 5
a more detailed and holistic representation of work [6] seconds or less to 5 minutes or more. depending on the
["1.
information available and modeling purposes. A model
might represent a group of people as a single agent, a
3. Components of a Brahms Simulation
useful heuristic in redesigning a work system. Common
Brahms has a multiagent language for describing objects and activities such as telephones and "'phone
agent and object activities. For a more detailed conversation" may be easily reused and adapted from
description of the language see [8] [4]. A Brahms other Brahms models. In general, Brahms models
stmulation of work practice has seven integrated represent work with much more detail than business
components: process models, but somewhat less detail (and far more
Agent Model: The group-agent membership broadly) than cognitive models (Figure 1).
hierarchy of the people in the work system. Groups may
be formal roles and' functions or based on location, Model Model Process
(inference) (activitiesl Mode!
interpersonal relations, interests, etc. _l,ll'tetlOrl $I
Object Model: The class-hierarchy of all the domain Figure1.RelationofBrahmsmodelstoothermodels
objects and artifacts, e.g., tools, desks, documents,
vehicles.
Considerable effort is devoted to modeling objects
Geography Model: The geographical areas in which (e.g., fax machines) and computer systems, with which
agents and objects are located, consisting of area- people interact to accomplish their work.
definitions (user-defined types of areas, such as
4. The Victoria Lunar Mission
buildings, rooms, and habitats) and areas (instances of
area-definitions}.
Victoria is the name of a proposed long-term semi-
Activity Model: The behavior of agents and objects
autonomous robotic mission to the South Pole region of
in terms of the activities they perform over time [9].
the Moon. The primary mission objective of Victoria is
Agent or object activities are mostly represented at the
to verify the presence of water ice and other volatiles
group-level or class-level respectively, but are also often
within permanently shadowed regions on the Moon. This
specific to agents and objects. Activities are inherited
will be accomplished by gathering the necessary lunar
and blended through a priority scheme.
data for analyzing the history of water and other
Timing Model: Constraints on when the activities in
volatiles on the Moon, and by implication in the inner
the activity model, can be performed, represented as
solar system. The Victoria team has decided to use a
preconditions of situation-action rules (called
high-speed semi-autonomous rover.
work'frames). Activities take time, as determined by the
One of the biggest constraints in any robotic mission
predefined duration of primitive actions. Workfi'ames is power consumption of the robot. In every acti:tty th_
can be interrupted and resumed, making the actual
rover uses energy, therefore the sequence of activities
length of an activity situation dependent.
for the rover is constrained by the amount of power
Knowledge Model: An agent's reasoning,
available to complete the sequence. Whcn the robot's
represented as forward-chaining production rules (called batteries are low, it needs to return to a sun-exposed spot
rhoughtlkames). Thoughtframes can be represented at
to recharge its batteries. During the Victoria mission the
group/class levels and inherited. Thoughtfi'ames take no rover will traverse into permanentl> dark regions on the
ttme Inquiry is modeled as a combination of activities
Moon. During these traverses the rover _,ill use its
Ie.g.. detecting intbrmatlon, communicating, and neutron detector instrument to detect hydrogen and the
reading/writing documents} and thoughtframes.
Sample Acquisition and Transt_r Mechantsm (SATM)
Perception is modeled as conditions attached to
f "_ / . ,,,,,_, . \ / \
, Humen Activity _ M, _ _ .4 / V_ofk r.ctlc.
_.. System ) q - _._.' : _ Simulation "
J observing work practice - Mode_rl - / observinq slmutated \
_ _ "" work oracllce \'-,_ j
/
Iformal _ ,
- o, /,,/4M:3,,oo yletds//
--\ / /
Figure2.CollaborativMeodelingProcess
to drill into the lunar surface and take surface samples to description (i.e. a conceptual model) of the work
be investigated using an array of science instruments. practice of a human activity system. For M1 we use the
The work system design problem is to configure the Compendium modeling method [I0].
mission operations so the robot's activities inside the Method M2 - formal model of the work practice:
permanent dark region are most efficient (i.e. consume The purpose of method M2 is to translate the conceptual
the least amount of energy). models from M1 into a Brahms model. The Brahms
modelers and the conceptual mode[ers do not necessarily
5. Collaborative Modeling have to be the same, and in fact, the skill sets for these
To develop a sirtiulation of the Victoria work system, twotypes of modelers are verydifferent.
Method M3 - simulation: Using method M3 a
we used a collaborative modeling approach with the
Brahms simulation of the formal model is developed, by
Victoria mission designers. First, we created informal
running the simulator with the tbrmal model as input.
static models of the activities of the people, robot, and
M3 isthe Brahms compile-simulate-debug method.
artifacts, the communication and geography, in informal
Method M4 - observing the simulation: The
design meetings and modeling sessions. Next, we
purpose of method M4 is to observe and investigate the
translated these descriptions into formal Brahms models,
work practice simulation output, and compare it with the
producing simulation results (including energy
measurements described below). The Brahms modeler actual human activity system. During this cycle the
went back to the mission design team and presented the actual objective of the work practice simulation project
result of the simulation. Together they discussed and is accomplished. Frequently, one modifies the model or
factual scenarios (e.g., for Victoria, the location of _ceira
represented more details of the work practice. The
the modeled geography} to perform a what-if analysis.
Brahms modeler implemented the new details into the
Thus, there is a modeling and simulation cycle
simulation. This Brahms modeling-debug-simulate cycle
between Mt. M2, M3 and M4. The methods have to be
continued until a sufficiently robust and interesting
closely integrated for the cycle to be et'fic_ent. For
design resulted•
example, to a certain extent we can automat;cally
Figure 2 describes our collaborative methodology for
developing first an intbrmal model and then a formal produce Brahms model components (M2) from ;nformal
s_mulation model of the Victoria work practice. Compendium descriptions (M I).
Method MI analyzing work practice: The
l
purpose of method MI is to create an informal static
Oiw'miI'lernf
II
_alle
Rover
Figure3.V'mtodwaorksyslem
[11]. They therefore specifically addressed the tbllowing
6. Mission Operations System Design
questions in the work system design for Victoria:
The work during in the Victoria mission will be 1. How will science data be gathered collaboratively
distributed over a number of human teams and the with the Earth-based science team, rover
Victoria rover. By virtue of being people's arms and teleoperator, and the rover onthe lunar surface?
2. How will science data be made available to the
eyes on the Moon, the teleoperated rover is more of an
science team?
assistant than a simple tool.
Figure 3 represents the work system elements and 3. What is the affect of a particular work _stem
their relative location during the Victoria mission. The design on the power consumption of the rover
Science Team consists of co-located sub-teams: the during a science traverse into a permanent dark
Science Operations Team (SOT), the Instrument crater?
S_ergy, Team (IST), and the Data Analysis and To answer these questions we de,,reloped a model or
Interpretation Team (DAIT). There are two other the activities of the teams, based on the description of a
supporting teams: The Data and Downlink Team (DDT) planned mission traverse. In the next sections we
and the Vehicle .and Spaceeral_ Operations Team describe the design of this work system through the
(VSOT). The teams communicate with the Victoria design of"the agent model, the object model, their
rover on the lunar surface using the Universal Space activity models, and the geographical model.
Network (USN), directly and via a lunar orbiter. The
6.1. Agent Model Design
flow of data from the rover will be dominated by
contextual data and science data. This data will come to Figure 4 shows the group membership hierarchy on
NASA,Ames via the Universal Space Network (USN) which the design of the work system is based. The
data connection and will be automatically conver_ed in agents in the model are the Earth-based human teams
near real-time to accessible data formats that can be and the Victoria rover, as shown in Figure 3.The teams
made available to the teams via data access and are represented as single agents, because at this moment
visualization applications. it is not possible to prescribe the composition and
Based on previous experience, the mission designers practices of each team in more detail. For example, the
hypothesized that many issues will affect the decision "'plan a command sequence" activity of the SOT
cycle of the science team, one of which is data overload represents the work of the whole team, while the
t A
Oml_ommun*¢alor T
_OUO
t _ j.*
I
/
¥_;¢_e md_Jc oo'ds"L
_roui)
',,, .¢
JI
I Omsm*J_da_l I
Figure4.V'_odaAgentModel
Table1.Fur_tionaalcdvitdyistributioonver_f_to_iateams
Oam _• V_le a_41 Rev_"
acid Sp_crdt
o1,4m_l
10=o Team
UIIKI_I 1Lm_lom_ ' ' ' i
ross ¢omow4fl¢mo_e
terence
om_mu_tu s_4m:o$ b,
_per=_n
Oo_,_,Hnk *O=s_
process ma_
2DIa _emmel
individual activities of each team member remain dynamically created data objects during the simulation.
unspecified. The Victoria rover is modeled as an agent The Victoria object model (Figure 5) includes classes
because it has activities, including primitive actions that for the science instruments on the rover and other
change the world, movements, and communications. objects contained in the rover, such as the carousel and
Table 1shows a possible distribution of the functions the battery. Furthermore, the model includes the data
over the Victoria teams [12]. Details of how different communicator class, which includes the objects for S-
teams collaborate to perform these fimctions constitute band and UHF communication. The model also includes
the work practice, as specified in the situation-action the sotb,vare systems that receive and convert the
rules (Brahms worliframes) of the different agents. mission data. A Brahms object represents the data
An example workframe for the SOT team forcreating visualization systems that present data to the Victoria
a command sequence for finding water ice is team. The Data and CoreSample classes allow
dynamically creating objects representtng spec=fic data
(paraphrased): When [ believe that there is a possibility
we can find water ice at the current location of the and lunar core samples during the simulation,
rover, then start the activity of finding water ice.
6.3. Geography Model Design
Generically, a workframe is of the tbrm: When (I believe
IX}*) Do {activity A. conclude a new belief and/or The geography model represents locations on Earth
fact }* and the Moon (Figure 69.The areas of interest on Earth
are Building244, where the Victoria teams and systems
6.2. Object Model Design are located, and UsnSatelliteLocation. where the
The object model consists of the classes and
instances of physical artifacts, as well the statically and
) DI[a {
I Co_ume¢ I
C , o3,-1. II I
_ I_tw*r'_ II I
Figure5.Vict_aOtectModel
I
I
I
Figure6.VictoriaGeographMyodel
UsnDishl satellite dish is located. Locations for the
7. Victoria Simulation Scenario
simulated scenario are represented on the Moon
ShadowEdgeOffSraterSNl represents the location the The case study selects one of the key _urface
rover at the start of the simulation (the shadow edge in activities, searching for water in permanently shadoa'ed
crater SNI). ShadowAreallnCraterSNl represents the craters:
area in the permanent shadowed $NI crater where the The rover has arrived at the shadow edge of crater
rover will perform a drilling activity. The LandingSite site number 1 The batter_" has been lit(cid:0)iv charged.
area is represented only for completeness. Based on the data analwts by the F.arth-ba,_t'd teams, oI
_o,N.,tt o,_lagt u_w'll_
Figure7.VictoriaRoverscenarioactivilies
the Clementine data available for the shadow edge area communication, and movement of each agent and object.
of crater site number I. the science team now decides After the model is developed and compiled, the Brahms
where to go into this crater and search for water ice. simulation engine executes the model. A relational
While the rover is traversing into the crater, it is taking database is created, including every simulation event. An
hydrogen measurements with the Neutron Spectrometer. end-user display tool called the .4gentViewer uses this
When the rover arrives at the assigned location within database to display all groups, classes, agents, objects,
this crater and it/inds hydrogen there, the science team and areas in a selectable tree view The end-user can
decides it should start drilling lOcm into the ,'urface select the agents and objects he she wants to investigate.
using the S,4TM. and collect a l.Occ lunar sample. The AgentViewer displays an activity time line of the
When the rover receives this command, it starts the agents and objects selected, optionally showing agent
drilling activity and finally deposits the sample into the and object communications. Using this view the end-
instrument carousel. user can investigate what happened dunng the
The rover uses two instruments in this scenario: the simulation. In the next three sections we explain the key
Neutron Spectrometer (to detect hydrogen--most likely behaviors during the simulation.
caused by water ice--within the first half meter of the
8.1. Data Uplink Activities
lunar surface below the rover) and the lunar surthce drill
tSample Acquisition and Transfer Mechanism -- The scenario starts with the Data Analys_s and
SATM). Interpretation Team (DAIT) retrieving the Clementine
The backbone of the simulation model consists of data image of the shadow edge area. where the rover is
three primary activities: Data uplink. Rover operations, located at the start of the scenario. They review this
and Uplink. These are described in the next section. image using their visualization system, represented in
the Brahms model as a VisualizationSystem object.
8. Simulation Results
According to the work practice, they do this without
The .,imulation provides visibility into the behavior anyone requesting that they look at the data.
,Jr the _ork system o_,er time, that _s, activities,
Figure8.Simulalionofdown_nkandsecond
uplinkcommandactivities
This means that the DA.IT needs to be know: l) the rain) into the crater to a specific location
location and situation el" the rover at all times, 2) (ShadowAreallnCraterSNI), and while driving it should
whether data is available and needs to be retric',ed, and be using itsneutron detectorinstrument re delcct
3) where and how they can retrieve data. hydrogen in the lunar surface This decision is
Once the DAIT has retrieved the images, it communicated to the Vehicle and Spacecraft OperatIons
communicates this to the Science Operations Team Team (VSOT), as well as to the DAIT After this
(SOT), and they collaboradvely analyze these images communication, the SOT waits tbr the rover's downhnk
(the AnalyzeRoverl'mages activity). When done, the data.
SOT plans the t_rst rover command sequence. According
8.2. Rover ,-_cdvity
to the scenario being simulated, the SOT decides that the
roverneeds t_drivel_)r:i,;pecntiedaxnountor"time(l5 -FhcVtct,_rLrt,;_er I_modeled as an agent, v_hereas
_hc neutron spectrometer and SATM instruments are detector and [ocation data, it retrtcvcs the dat:l l'rtlm rhc
luodclcd as ,_eparate science instrument objects VisualizationSystem object (the ,tctlv_t_cs
c_mramcd in the rover agent. [n the scenario model, the RctrieveNeutronData, lntcrpretNcutronData, and
Ncumm Spectrometer object is active and creates a FindRoverLocationData)
IlydrogcnD;,ta_l object conta.ning the hydrogen data Next, the DAIT commumcates their r_ndings to the
that _s,,cnt to Earth while the VictoriaRovcr _straversing SOT In the example ,;ccnario. thedata ,,uggest that the
to a permanently shadowed area within the crater SNI rover has found hydrogen m Shado_-X.rca[ InCratcrSnl
The rover then waits tbr the next command sequence Given this finding, the SOT quickly determines the next
from Earth. During this time the teams on Earth are command sequence for the rover and communicates this
analyzing the hydrogen data and deciding what to do decision to the VSOT (see 'Next Ro_cr Command
next. In the Uplink activity, the rover is given the Decision" in Figure 8).
command to search tbr water ice in the permanent dark The communication informs the VSOT to transmit
area This triggers a traverse, during which the SATM the command sequence to the Vtctor:aRo,,er (see
instrument will start the drilling activity. "Create command sequence" in Figure ;3_.The command
To collect a sample the SATM has to 1) lower its sequence tells the VictortaRover to start the
augur to the surface, 2) drill to the depth given as part of SearchForWaterlcelnPermanentDark.-krea activity. It
the command by the SOT (in this scenario the command also tells the VictoriaRover that its sub-activity is to
says to take a 1.0cc sample at 10¢m depth), 3) open the perform the Drilling, Activity. Parameters indicate how
sample cavity door, 4) continue to drill to collect the deep to drill and how big a sample to collect at that
sample, 5) close the sample door when done, 6) retract depth. Figure 8 shows part of this second uplink process.
the drill from the surface, and 7) deposit the collected The duration of the downlink and second uplink
sample on the instrument carousel. (see "Drill 10¢m into processes determine the duration of the second
surface and take lcc sample" in Figure 7). DoNothing activity of the VictoriaRover, simulating the
In the Brahms model, the Augur object creates the time the rover is waiting for the Victoria science team to
LunarSample_l object as part of its activity to capture decide the next command sequence (see "'Waiting for
the lunar sample, after opening the sample door and Command from Science Team" in Figure 8).
continuing the drilling to collect the 1.0cc sample. The
activity times for drilling into the surface are 8.4. Modeling Energy Consumption of Rover
dynamically derived during the simulation. To calculate the total energy used by the rover, we
8.3. Downlink Activity need to represent in the model the energy, needed tbr
each subsystem during a rover activity lsee equation
When the rover detects hydrogen in (I)). The total power consumption of the rover during
ShadowArea I[nCraterSNl the downlink process starts the scenario can then be calculated (equation 12)).
(represented by the Brahms AgentViewer in Figure 8).
The VictoriaRover agent contains the S-BandMGA
object, which represents the S-Band transmitter on the To_dPowerConsumpncn=i=_oE,cn' 11)
rover. The VictoriaRover creates a data object with a)
the current rover location information and b) the
hydrogen data. This data object is then communicated to = amaor=¢=Prover(/ldl 12)
Earth, via the UsnDishl object. The Usn.Dishl object
The energy consumption for every rover activity
communicates this data to the DataConversionSystem,
located at NASA Ames. during the simulation of the scenario is shown in Figure
As can be seen in Figure 8 (see "Downlink process"), 9. In particular, the energy the rover uses during the
the DataConversionSystem performs two conversion Waiting activity (see "waiting for command from
activities, one for the hydrogen data and one tbr the science team" in Figure 8) is defined by the energy."
needed for Thermal Protection during driving -
location data from the rover The work system design
requires that the data conversion system interact with the Command and Data Handling ,htring driving. While the
visualization system without human intervention. rover is standing still and "'doing noth,ng.'" it consumes
power for its thermal protection and its commanding and
When the VisualizationSystem receives the newly
converted data, the system alerts the DAIT. A member data handling tbr its subsystems, such as its processor
board.
of the DAIT monitors the VisualizationSystem while in
the activity WatchForDownlink (see "'Detect, retrieve, Besides the power left to use after the scenario.
another interesting vanabte is the energy, usage rate b,,
interpret and communicate data" in Figure 8). When the
the rover (equation (3U
DAIT agent detects that there is newly available neutron
"D ,
t'llcr_vR.dc =l',datP,_wcr Phallcry(_larolr"traversc-_ (3) [21 F E. Emery :lnd I! L. Tr_,,,t."S_c=o-rcchmcal
Fh=sequatl_m -;howsthat given rhcenergy usedtnthe Systems." in _/anclgemt'rtt ,_'t'tdnt t'_. _h,h'l_
and Techmque_. f" W Churchna_m. lid.
sccnarm- -drive 900m into the crater, and rake one 10cc
London: Pergamon, Ic)60
';ample ,tt I()cm depth--with the current work system
[31 J. Grecnbaurn and M. Kyng. "Design ,it Wurk
design, the robot has used almost a third of its power:
Cooperative design of cornputcr systcrns."
Ener_yRatefdrtllinE inpermanent dark crater) -_O.30 Hillsdale. N.J'.:Lawrence Erlbattm, l')t}l
[41 M. Sierhuis, "Modeling and Simulating Work
1 Practice; Brahms A muhiagcnt modeling and
simulation language tbr work system Jnalysns
and desing," in Sot'tel Science and htti_rntuttc'_"
(SWI). Amsterdam, The Netherlands:.
University of Amsterdam, SIKS D_sserratnon
]= Series No. 2001-10, 2001, pp. 350.
[5] M. Sierhuis, W. ,1'.Clancey, R. v. Hoof, and R.
d. Hoog, "Modeling and Simulating Human
!
t /./// / Activity," presented at AAAI FallSymposium
I. on Simulating Human Agents. North Falmouth,
ii MA, 2000.
[6] W. 1. Clancey, P. Sachs, M. Sierhuis. and R.
van Hoof, "Brahms: Simulating practice tbr
work systems design," International Journal on
Human-Computer Studies, vol. 49, pp. 83i-
Figure9.Roverenergyusedinhigh-levealctivitiefsrom 865, 1998.
simulatiohnis[o_database
[7] M. Sierhuis and W L Clancey, "Knowledge.
Practice, Activities, and People," .4.4.-I1Spring
This variable represents the rover power Symposium, B. Gaines, Ed Stanford
consumption, a measure of the effectiveness of the work
Universi_, CA.: Proceedings of AAAI Spring
system design, which can be used to compare different
Symposium on Artificial Intelligence in
work configurations for the modeled scenario.
Knowledge Management, 1997. pp. 142-148.
9. Conclusions [8] R. van Hoof and M. Sierhuis, "Brahms
Language Reference," NASA/Ames Research
In this paper we described how Brahms was used to Center, 2000.
design the mission operations work system for the [9] W. Clancey, J., Situated Cognition. On Human
Victoria mission.
Knowledge and Computer Representations:
The Brahms simulation showed the impact of the Cambridge University Press, 1997.
work practice of Earth-based teams on activities and [I0] A. Selvin, S. B. Shum, M. Sierhuis, J. Conklin,
energy consumption of the rover. The simulation allows B. Zimmermarm, C. Palus. W. Drath. D. Horth,
mission designers to compare different work system J. Domingue, E. Motta, and G. Li,
designs before critical mission decisions have been "Compendium: Making Meetings into
implemented.
Knowledge Events," presented at Knowledge
We have shownan instance of a collaborative work Technologies 2001, Austin, TX, 2001.
system design methodology incorporating simulation. [11] G. Thomas, M. Reagan, E. A. Bertis lIl. N.
The Brahms simulation framework provided guidance to Cabrol, and A. Rathe, "Analysis of _cience.
mission and robot designers, replacing a spreadsheet team activities during the 199',) Marsokhod
approach by a more transparent and flexible multiagent rover field experiment: Implications tbr
representation. automated planetary surface exploration." The
University of iowa, Io_vaCity 19')9
I0. References
[12] S.D. Wall and K. VV Ledbcttcr, De_-tgn of
[11 T. H. Davenport, Process Innovation. Re- Mission Operations Svstent_" l_Jr Scientific
engineering Work through In]brraation Remote Sensing. London: Taylor & Francis.
Technology Boston. MA: Harvard Business 1991.
School Press, 1993.
I0